ABE15 Lars Roost: Scaling agile at LEGO
Agile By Example 2015 · 25 min
Vídeo: https://youtu.be/RPhCagS9II8
Transcrição automática do YouTube (inglês, legenda gerada automaticamente). Timestamps em [mm:ss]. Erros de reconhecimento de fala não foram corrigidos.

[0:04] my name is lass I work with Lego uh more
[0:06] specifically I work with our consumer
[0:08] marketing agency never thought I going
[0:10] to say that when I graduated as an
[0:12] engineer um I work with AUD digital
[0:15] experience platform which is the part um
[0:18] that we buildt all our Global Marketing
[0:21] um on so that's websites apps and even
[0:24] some of our games tied into that what
[0:26] I'm going to talk to you about uh I
[0:28] found out this morning is actually round
[0:30] about so my official title is a program
[0:32] manager uh which means that I'm charge
[0:35] of budgets timelines and that kind of
[0:37] stuff all the traditional uh management
[0:39] kind of site on on running big programs
[0:41] running projects uh I do like to
[0:44] consider myself uh a bit of an agile
[0:46] Alchemist as well um which is the reason
[0:49] I'm here today um I'm so fortunate that
[0:51] uh I have an organization of about 140
[0:53] people um that I'm allowed to run quite
[0:56] some experience uh on and one of those
[1:00] experiments is the one I'm talking about
[1:03] today but before I'm talking about it
[1:05] awesome thank you um we're already there
[1:09] um but before I going to talk about it
[1:11] um I'm going to take you back about a
[1:13] year um run through a short introduction
[1:16] and then take you right with me into the
[1:19] mud going through a lot of pictures of
[1:22] what scaling agile at Lego consumer
[1:24] marketing is actually all about um cool
[1:28] so it started a year ago um when I
[1:31] started my current position and I was
[1:33] asked the question so how do we scale
[1:35] that agile thing um we had it back then
[1:37] about 14 development teams running scrum
[1:40] mban uh we had a demand exceeding
[1:42] capacity by a couple of four to three
[1:45] times and we were looking into a market
[1:48] that was changing ever faster and we
[1:50] needed to react faster and faster on
[1:53] Trends um so so how do we do that my
[1:57] first thought is well we don't we don't
[1:59] scale um let's keep things small
[2:01] decoupled only scale that the stuff that
[2:04] really really needs to um but I pretty
[2:06] soon found out that that wasn't really
[2:08] an option um the technology stack wasn't
[2:11] as decoupled as it should have been um
[2:13] people were delivering a lot of products
[2:16] that had all kinds of dependencies to
[2:18] each other so team were one Sprint
[2:20] working together on one product in
[2:22] another Sprint working together on a
[2:23] different product and there were all
[2:25] kinds of both human and Technical
[2:28] dependencies all over the place
[2:30] so the real question was is how do we
[2:33] scale agile when we're already big and
[2:35] messy when we're in the middle of it um
[2:38] I thought that came to mind is what kind
[2:39] of what that kit up there does uh Smash
[2:42] It uh let's start over uh let's build it
[2:45] from the bottom but obviously uh in a
[2:48] global company you can't do it um you
[2:50] need to keep delivering um marketing
[2:53] campaigns you need to keep the wheels
[2:55] running uh about 70% of the models that
[2:58] Lego brings out are new every year and
[3:01] they don't sell themselves well some of
[3:02] them do but they do need some marketing
[3:05] material as well
[3:07] so that's the solution um big room
[3:10] planning we put 140 people in a room for
[3:14] two days and we run a lot of experiments
[3:16] that's our roundabout um and that's what
[3:19] I wanted to talk to you about today um
[3:22] the pattern of using big room planting
[3:24] and all the small micro patterns within
[3:27] that big
[3:28] pattern so it started for us with um the
[3:32] belief that we need to focus on the
[3:34] basics we need to amplify simple things
[3:38] and we need to amplify uh collaboration
[3:41] at the lowest levels so when you think
[3:43] about scaling stuff you it typically
[3:46] introduces a lot of overhead introduce
[3:48] the need for coordination the need for
[3:51] aligning priorities um all kinds of
[3:54] stuff that that just drags uh teams down
[3:57] compared to when they were just on their
[3:59] own
[4:01] so there it is um let's scale that was
[4:05] the task and there's there's definitely
[4:07] no s shortage of Frameworks uh out there
[4:11] to to choose from and these are just a
[4:13] couple of them um the one we chose
[4:16] probably comes as Sur prise I'm
[4:18] expecting at least a couple of boo now
[4:21] um because uh a scale Edge framework if
[4:24] you read about it on the net uh internet
[4:27] you'll probably found uh dozens of rants
[4:31] going on about the bad things about this
[4:33] and I I do believe that it's it's a
[4:35] tricky one um but it's it's basically uh
[4:40] saying that safe is bad bad is the same
[4:42] thing as saying that a certain cookbook
[4:44] is bad because that's really what safe
[4:47] is it's a big heavy cookbook with a lot
[4:49] of recipes um and it does lead you in
[4:52] the way of thinking uh I probably cook
[4:55] all my meat this way but you can go
[4:58] about it in a different way you can
[5:00] get inspired by some of the recipes and
[5:02] and that's what we did so the situation
[5:04] we faced is basically we we we had that
[5:06] lower part here we had scrum teams
[5:09] running on sync Sprint Cadence um we had
[5:12] a lot of them we had that huge budget
[5:14] process that even was bit canand
[5:17] inspired that covered all of Lego so not
[5:19] just us living in a in a software
[5:21] development world but also all the
[5:24] manufacturing part impacted by a yearly
[5:26] yes yearly budget cycle um so we needed
[5:29] something in between we needed some
[5:32] program levish thing that would
[5:35] interface the agile software development
[5:39] that we were doing in the team and the
[5:41] the world around it that didn't really
[5:43] work in an agile way that didn't know
[5:46] what software development or the
[5:48] principle of of of Agile development
[5:50] really are um so what we liked about the
[5:54] program level is basically the concept
[5:56] of having the restra engineer um
[5:59] basically a uber scrum Master somebody
[6:01] who can speak of behalf uh of a team of
[6:04] teams somebody who's in charge of
[6:07] processes that are not in the team but
[6:09] shared across and and that really worked
[6:12] for us uh we also thought but that gives
[6:15] us a chance to dress up I'll show you
[6:17] why later um but that's basically what
[6:21] we took from it so we didn't go uh all
[6:23] in we didn't introduce all of safe all
[6:25] the roles all the many many things that
[6:28] aren't safe um we we stick to that and
[6:31] the story of to our top management say
[6:33] we got it covered um we have a solution
[6:36] um just let us let us do it our way um I
[6:40] start with one thing first um and that's
[6:43] the shared Cadence it's really more
[6:45] important than you used to think um now
[6:49] safe works with a with a two-tier
[6:51] planning Cadence um you have the Sprints
[6:54] as you would have in a normal scrum
[6:56] setup and we had those and we kept those
[6:59] uh running on on two weekly uh Sprint
[7:02] Cadence and we introduced program
[7:04] Cadence on top of that and 8we Cadence
[7:07] um that would govern the timing of our
[7:09] planning uh it's pretty important to say
[7:12] here that this governs the timing of
[7:13] planning not the timing of releases so
[7:17] we used to and we still do uh run a
[7:19] continuous delivery organization so
[7:21] stuff can be delivered at any time um we
[7:24] don't buffer stuff up to the end of
[7:26] Sprints or to the end of program
[7:27] increment we do release it when it's
[7:29] ready
[7:30] um but we plan on a fixed Cadence that
[7:33] allows us to take uh care of a lot of
[7:36] the complexity that this big
[7:38] organization of now about 20 teams um
[7:43] gives us so traditionally um safe talks
[7:46] about an Agile Release Train a team of
[7:49] teams working together on one product uh
[7:52] delivering a a coherent experience um
[7:55] and that's not really our case we're
[7:57] delivering a lot of different marketings
[7:58] campaigns on a l of different marketing
[8:01] challenge uh channels um but they're
[8:06] somehow interconnected and that's why we
[8:08] chose let's let's have this not for as
[8:10] our Release Train let's have for
[8:11] everybody so let's all development teams
[8:13] run them on a single Cadence have them
[8:16] all sit together maybe they're only
[8:17] Loosely coupled but maybe they'll be uh
[8:20] more tightly coupled uh in a Sprint or
[8:23] in 10 um so let's just start by having
[8:25] them all together and see how it goes
[8:28] and then came big room plan
[8:30] well not really uh the first time it
[8:32] looked like this um we had a room that
[8:35] was too small to fit everybody in um it
[8:38] was really really noisy what kind of a
[8:40] humming sound like in here um we even
[8:43] collided with the release dates of some
[8:45] of the teams that had to leave in the
[8:47] middle of of the planning session and
[8:49] everybody left with a headache in the
[8:50] first day because it was simply
[8:52] impossible to to sit in in this tight
[8:55] room and do a Sprint planning kind of
[8:58] activity
[9:00] um but we had fun hats uh for all roles
[9:04] um so um let's let's let's look at how
[9:09] that worked it actually worked quite
[9:12] well so despite of all the headache
[9:14] despite of the noise and the frustration
[9:15] about the surroundings um instantly the
[9:18] team felt that this was something we
[9:20] needed to work continue work with this
[9:22] was something that that put the
[9:24] communication face to face um people
[9:27] were talking with other teams uh
[9:29] instantly because they were just sitting
[9:31] next on the next table um we would have
[9:34] the stakeholders in the meeting already
[9:36] from the first time so so a lot of
[9:38] communication what happened face to face
[9:40] instead of email or phone and on top of
[9:44] that when when all based on all that uh
[9:47] communication the teams came up with the
[9:49] first plans um so this would be a plan
[9:52] high level for for Sprints um they would
[9:54] receive instant feedback so they could
[9:57] say well D all team can you can you do
[10:00] this at the second Sprint yeah we can do
[10:02] that ah and can you do this now we can
[10:04] do that we do bit later um and they
[10:07] would receive instant feedback based on
[10:09] management uh from management so they
[10:11] could show their plans and say we're
[10:13] probably going to do this in the next
[10:14] two months and management could say yeah
[10:16] that's awesome oh not not so awesome so
[10:19] let me yeah ties back to the what we do
[10:23] let me take you
[10:24] through um how that actually looked like
[10:27] looks like so this is an overall genda
[10:29] for the last peer plan that we did uh a
[10:32] month ago um main takeaways from this
[10:35] slide all the blue areas are
[10:37] communication from the stage to the
[10:39] audience kind of like this all the
[10:41] yellow uh topics are uh Team breakouts
[10:44] that's where the team actually sit
[10:46] together at a uh at a table and do uh
[10:50] Sprint planning kind of work um and the
[10:54] red one is a management only I'll get
[10:55] back to that I'll keep it for your
[10:57] reference up here so you know where I am
[10:59] it typically
[11:01] starts with feedback feedback on the
[11:04] last Pi um we tell the teams how we
[11:07] think they did um did they deliver what
[11:11] they what we hope they would um both
[11:14] from management yeah you did awesome or
[11:16] no we didn't do what that customer
[11:18] wanted um and a bit from me some hard
[11:22] facts um a lot of different kind of
[11:25] information that we think are useful for
[11:26] the teams to to have in mind and to
[11:29] think about when they plan the upcoming
[11:32] program
[11:33] increment um we do have a system demo
[11:36] that's a Sprint Dem for everybody just
[11:39] before um we go into planning but this
[11:41] is kind of the gist of it and and
[11:44] focused on what the teams need going
[11:46] forward um there's also an celebration
[11:49] element in this let's see with all the
[11:51] technical luck we had onto far at this
[11:53] actually place um you may cover your
[11:55] ears I have no idea how the sound
[11:56] settings are um
[12:00] they are not very
[12:02] loud anyway you would be hearing some
[12:05] music and a lot of cheering and clapping
[12:07] um it's really an awesome when you get
[12:10] that many people in a room and they they
[12:13] really concentrated you get something
[12:14] you need something to get them started
[12:16] and this really did did it for us um
[12:19] it's it's just something to get people
[12:22] started
[12:24] whoop interesting
[12:31] there we go this is followed by a series
[12:34] of lightning talks um these cover a bit
[12:37] of everything um the main purpose is to
[12:41] inspire the teams and to guide the teams
[12:45] uh on their way into into planning um so
[12:47] there'll be something about high level
[12:50] priorities um these come either as uh
[12:53] General guidelines for everybody we
[12:55] really really want to focus on some
[12:58] architectural uh aspect that shared for
[13:00] everybody and they also come in the form
[13:03] of high level priorities from our main
[13:05] customers that we think everybody should
[13:07] know about um of course the product
[13:10] owners for each team have additional
[13:12] details for that are relevant only for
[13:14] for one team um but this is kind of the
[13:17] the main uh stories um we also put a lot
[13:20] of effort into finding topics we think
[13:23] the teams think about so something about
[13:26] uh digital child safety that's of course
[13:28] pretty important when we do marketing
[13:30] for children um Global insights which is
[13:33] um the central unit and Lego that
[13:35] collects all the insights about what
[13:37] works what for kids what are they
[13:39] interested in um giving some perspective
[13:42] for that to our development teams
[13:43] changes and Privacy Law or even
[13:46] something internal like a hackathon we
[13:48] did some months ago so all kinds of
[13:50] information that we want to uh give the
[13:52] team so that they can plan better uh and
[13:54] can deliver better
[13:57] products shortly after this
[13:59] first half of day uh when we had all
[14:01] these kind of um topics shared to teams
[14:05] we collect feedback instantly so we have
[14:07] feedback boards uh on the way out of the
[14:09] room where we ask the teams uh uh to to
[14:12] dot vote what they like what they didn't
[14:14] like we ask for um uh concrete uh
[14:18] feedback as well so we try constantly to
[14:20] evolve this
[14:22] format and then after short launch break
[14:25] they start to plan so you'll notice uh
[14:27] color t-shirts teams are really getting
[14:30] into trying to to to make team um team
[14:35] team identities um and basically the
[14:37] setup is one big Sports Hall um with a
[14:40] lot of tables whiteboards and the teams
[14:43] working um with themselves trying to
[14:46] plan but also working with each other
[14:49] kind of visiting each other talking
[14:51] about uh technical aspects talking about
[14:54] dependencies um so it's it's really a
[14:57] high energy uh setup um you can't keep
[15:02] concentrating for so long so of course U
[15:04] we ask the teams that they can spend the
[15:07] time as they see fit some of them choose
[15:09] to build a bit of Lego um but we do ask
[15:12] of all of them to stay in the room
[15:14] that's really important for the for the
[15:15] teams in such an event to know that even
[15:18] though you don't uh work actively on
[15:20] planning right now it's really important
[15:22] that you stay here because all the
[15:24] others some of the others may do and
[15:26] they may need you uh some along their
[15:29] road towards a a plan for the next two
[15:32] months
[15:35] um what they plan as uh pulled off a
[15:38] feature backlog now as I told you we're
[15:40] doing a bit of different uh experiences
[15:43] and products during such a session a lot
[15:45] of different things are planned um some
[15:47] of the teams work on on a shared backlog
[15:49] so uh they pulled features of a feature
[15:52] backlog that we prepared um this was how
[15:55] it looked at the second or third
[15:57] planning uh today it's a it's a digital
[16:00] one um we're trying to constantly also
[16:02] evolve that part um trying both online
[16:05] and offline versions currently it's it's
[16:07] a the tool that we use for scrum um I'm
[16:10] sure other tools would support us as
[16:12] well um just to visualize and collect
[16:15] data on um how we're doing um and we
[16:19] actually use this afterwards to follow
[16:22] up like how are we doing with planning
[16:24] are the teams progressing as we think
[16:26] they will um are some of them over
[16:28] committ committing are they underc
[16:30] committing um and in general how how
[16:32] does the the backlog in general look
[16:35] have they pulled as much as we hoped
[16:38] um another thing to synchronize around
[16:41] are the dependencies I think that's
[16:42] really important um thank you so having
[16:46] 20 teams with a lot of dependencies uh
[16:49] can really be tricky uh we saw a lot of
[16:52] uh delays early on uh because
[16:54] dependencies between teams simply
[16:56] weren't caught um and we couldn't simply
[17:00] we couldn't centralize it so instead we
[17:02] chose a centralized tool for
[17:04] decentralized dependency mapping so
[17:06] whenever team discovers a dependency to
[17:10] another team it takes that team up to
[17:12] the board they just put posters on it
[17:14] connect them with a red line um these
[17:17] are Sprints these are all the teams um
[17:21] and over the day course of these two
[17:24] days this board grows into something
[17:26] like this um so we have an overview of
[17:30] all the dependencies we can instantly
[17:32] see patterns so you'll notice here's one
[17:34] team that's really has a lot of
[17:36] dependencies to it and there's some here
[17:37] that pretty much work independently of
[17:40] everybody else so I as a program manager
[17:42] also know going into the program uh
[17:45] increment where to focus my energy what
[17:47] to watch and we do that also during the
[17:49] scrum of scrums where we bring this
[17:51] sport back with us uh at the office and
[17:53] have the scr scrums in front of it kind
[17:55] of followup is this still progressing as
[17:57] we think are there problems turning up
[17:59] along the
[18:01] way another tool we use are risk boards
[18:05] um so you you know you can roam the
[18:07] risks you can resolve own accept and
[18:09] mitigate risks um but in my experience
[18:12] scrum teams aren't using risks uh
[18:16] actively as not as as actively as I as a
[18:19] former project manager would like them
[18:20] to anyway um so so we use these boards
[18:23] and we introduce them either for project
[18:25] that are shared between more teams or or
[18:28] really projects um and we asked the team
[18:31] to put up all the risks along uh the day
[18:34] when they plan that they they they
[18:36] uncover and we ask them to talk with
[18:38] their manager about it instantly so that
[18:41] we we try to not buffer up risks to some
[18:44] risk session but address them as soon as
[18:47] we see them um and also kind of
[18:49] visualize it and have it for followup
[18:53] um
[18:55] actually no let's have this first um so
[18:58] once the team have uh done about two 2
[19:01] and a half hours of of breakouts and
[19:03] have a first draft plan of what they
[19:05] think they'll do on a high level for the
[19:07] next two months uh we'll have a we'll
[19:10] have a draft review safe talks about
[19:12] this is a big session where everybody
[19:13] presents to everybody we tried that that
[19:15] was really painful because not
[19:17] everybody's interested in was what
[19:19] everybody else uh is doing so instead we
[19:21] have this format where basically we have
[19:23] four rounds where the team ex uh leaves
[19:26] one behind who presents their own plan
[19:28] and everybody else can go to a team if
[19:31] they're choosing and listen in on what
[19:32] they think they will do uh during the
[19:35] next two months it's really a again a
[19:38] great tool for decentralizing uh
[19:40] knowledge sharing so we let the people
[19:41] choose themselves what kind of knowledge
[19:43] they need um and and it it it actually
[19:48] works really really well uh the first
[19:50] you typically see that they they you use
[19:52] a bit of more time the first two times
[19:55] um because they have a couple of teams
[19:57] that they are closely interacting with
[19:59] and then the last two times with 5
[20:00] minutes is is sufficient for just maybe
[20:02] just the listening in on a team that you
[20:04] may or may not think that uh um will be
[20:07] relevant for you after this uh all the
[20:11] team members uh leave for the day it's
[20:13] it's it's a normal work day so behind
[20:16] stays the management team and that's
[20:18] everybody from the line manager for the
[20:20] teams up until the senior manager where
[20:23] is Marco he I think he's over here
[20:24] somewhere behind there um who's kind of
[20:28] uh the senior director for for all of
[20:30] the digital marketing um and what we do
[20:32] at this evening event is we go through
[20:35] all the
[20:36] backlogs um the product manager kind of
[20:39] going through I am I confident that we
[20:41] can do this um are there any concerns
[20:44] with what the teams have pulled do are
[20:46] we missing something or we even
[20:47] delivering more than we expected and
[20:49] also going through all the risk boards
[20:52] uh looking at are there any risks that
[20:53] need escalation and we address this at
[20:56] uh the evening of the first day so so
[20:58] that when the second day starts we can
[21:00] start the second day with instant
[21:02] feedback from Management on are we
[21:04] satisfied uh are we see do we see any
[21:06] problems with what you uh uh with your
[21:09] Draft plans for the next two months are
[21:11] there any directions we need to to
[21:13] change um are there any risks that we
[21:15] already have addressed or that somebody
[21:17] in management has taken ownership uh
[21:20] off and we send then the teams off to
[21:23] another couple of hours of Team
[21:25] breakouts uh my experience is they they
[21:27] usually are done relatively early in
[21:30] this block but again some really do use
[21:33] all the time and uh and all the others
[21:35] stay stay there they start to develop or
[21:38] they start
[21:39] [Music]
[21:40] to uh answer emails uh spend their time
[21:43] as as they see fit um again we we're
[21:46] hiring smart people so we shouldn't get
[21:47] in the way of of doing what they think
[21:49] is right um we'll end the event with a
[21:53] final plan review uh currently we are
[21:56] still using the format that everybody
[21:57] presents to every body and we also
[22:00] giving feedback from the product
[22:03] managers and Senior Management on what
[22:04] they heard instantly um it's it's maybe
[22:06] the weak part I think uh of this event
[22:09] and we we're thinking about ways to
[22:11] change this and of course then we end
[22:13] with the confidence road so still we we
[22:16] really try to create a transparency into
[22:19] risks into plans um and ask the teams to
[22:23] to give us feedback on whether they
[22:24] think this will work or not um the event
[22:29] ends with a
[22:30] retrospective um the teams uh do
[22:33] retrospectives uh for for themselves and
[22:35] then have the scrum Master presented to
[22:38] us at a a shared session and we kind of
[22:42] both ask for how how is this good for
[22:44] you does this work for you um as you can
[22:46] see this one is from last time General
[22:48] we're getting really positive feedback
[22:49] from the teams but we do still have uh
[22:52] some ones and twos um but we ask them
[22:56] well so what should we change and what
[22:58] should we do more of so so this format
[23:00] really changes from every time
[23:05] um so we think actually that we we uh
[23:09] managed to get more collaboration uh in
[23:12] a much simpler way than we had before um
[23:15] and it did a lot of other things for it
[23:17] uh that I unfortunately don't have time
[23:18] to cover in 30 minutes so it's it's
[23:21] really a driver of momentum for us um
[23:24] having these big events every second
[23:26] month drives commitment from the
[23:28] business side drives commitment from our
[23:29] own side um it creates a lot of
[23:32] awareness of that we're undergoing a
[23:34] change process and that change is
[23:35] constant so so it's not small
[23:38] incremental changes but it's it's in
[23:40] your face change um also towards the
[23:42] business around us um and as a side note
[23:45] it actually turns out to be quite a nice
[23:48] buffer um between operating in an it uh
[23:52] Department waterfall world and our uh
[23:55] marketing agile soft development world
[23:58] where the two-month Cadence kind of
[23:59] seems to be a reasonable uh Middle
[24:02] Ground at least uh at least our company
[24:05] so
[24:06] um let me finish with this slide on a
[24:08] key takeaways so we really think that
[24:10] big panning can be a a quite useful
[24:13] pattern to use it's certainly not the
[24:15] right one for everybody it was or is the
[24:18] right one for us um it gives us more
[24:20] collaboration uh it gives us more
[24:22] alignment on when we plan and what we
[24:24] plan and how we plan um it gives us a
[24:27] lot more transparency into the risks and
[24:30] the dependencies that we are working
[24:32] with um not more just the same as before
[24:35] but now we know them um and it gives us
[24:38] a lot uh a lot of other nice things so
[24:42] um that's it for me thank you for your
[24:45] time